“Goodbye, $165,000 Tech Jobs. Student Coders Seek Work at Chipotle.”
“Skip the Computer Science Degree.”
When anyone can ask an AI, in plain English, to build a website or write a working program, it is tempting to ask: is programming really dead?
Neuroscience suggests otherwise
Programming is much more than translating an idea into lines of code. In fact, the human brain relies primarily on reasoning systems, rather than the language system, when understanding programs. Programming requires us to break problems apart, construct algorithms, anticipate how those algorithms will behave, and reason backwards when something goes wrong. With experience, programmers also develop ways of representing and thinking about algorithms that novices do not automatically acquire.
Human input is required to fix the last “30%”
AI makes it dramatically easier to turn an idea into a working prototype. But that creates what Addy Osmani has called the “70% problem”: getting the first 70% can be surprisingly easy; getting through the last 30% that separates a superficially convincing prototype from reliable, deployable software can be much harder. That last mile requires understanding the problem, spotting failures, handling edge cases, and knowing what needs to be fixed in the first place.
AI therefore offers an exciting new entry point into programming. But computational thinking does not simply emerge from casually prompting an AI.
Programming is changing.
It is not disappearing.
And it remains a skill that needs to be learned.
Our new commentary, Programming education in the AI era, is now out in Nature Reviews Psychology.
